Provides a read-only view to the model's output tensor(s). Retains ownership of object.

Override existing hyperparameter values, parsing new values from a string. See parse_values for more detail on the allowed format for values.

Returns the list of all local per-replica values contained in `value`. DEPRECATED: Please use `experimental_local_results` instead. Note: Th

Returning the inputs as a list.

Returns the inputs handle shapes and types, for the resource tensor output at index <idx>. Returns NULL if the shape and types were not available.

Returns the mode in which the file was opened.

Returns a tensor shape for the specified format and dimension sizes. Works for both 2D and 3D operations. The output shapes are as follows: FORMAT_NHW

Add operations to minimize `loss` by updating `var_list`. This method simply combines calls `compute_gradients()` and `apply_gradients()`. If

Minimize `loss` by updating `var_list`. This method simply computes gradient using `tf.GradientTape` and calls `apply_gradients()`. If you wa

Return a unique operation name for `name`. Note: You rarely need to call `unique_name()` directly. Most of the time you just need to create

Formats hipError_t to output prettified values into a log stream. Error summaries taken from: TODO(leary) switch to cuGetErrorName when updated rocm.

A float val -> Tensor<float>

Op converters (e.g. ConvertReshape) need to access the TRT network in order to add TRT layers.

The output tensors of this subgraph view.

Compute the number of elements in this table. Args: name: A name for the operation (optional). Returns: A scalar tensor containi

Number of elements currently held by this TopN object. This will be no greater than 'limit' passed to the constructor.

Test-only.

Returns the number of parameters for this computation.

Get all the valid tests configs to run. Returns: all the valid test configs as tuples of data_format and use_gpu.

Prepend the current test case's working temporary directory to <prefix>

Return the instruction which defines this HloValue.

Construct a `DeviceSpec` from a string. Args: spec: a string of the form /job:<name>/replica:<id>/task:<id>/device:CPU:<id> or

Returns the value of `key` if it exists, else `default`.

Creates a constant tensor. The resulting tensor is populated with values of type `dtype`, as specified by arguments `value` and (optionally) `sha

Returns a hashable reference object to this Tensor. Warning: Experimental API that could be changed or removed. The primary usecase for this

Returns whether any errors have occurred while entraining work for this stream.

How many elements are in the array with this shape.